svctm

svctm is a performance metric used in some UNIX-like operating systems to describe the average time a storage device spends actively servicing an I/O request. The term is commonly read as “service time.” In practice, svctm reflects the time the device spends on the work of executing a request, excluding time spent waiting in queues.

The availability and exact meaning of svctm vary by platform. It has appeared in the output of older System V and Solaris iostat tools and is not universally present in modern Linux tooling. When reported, svctm is typically expressed in units such as milliseconds or, on some systems, jiffies or microseconds. Because different systems may define and calculate it differently, svctm should be interpreted in the context of the specific OS documentation and version.

Relation to other metrics is important for interpretation. Svctm is often shown alongside throughput figures (transactions per second) and bandwidth, and it is related to, but distinct from, the average wait time. In some contexts, a high svctm indicates that the device is taking longer to service requests, which can point to contention, slower hardware, or suboptimal I/O scheduling. Conversely, a very low svctm with high utilization may reflect efficient servicing under heavy load.

Limitations include lack of standardization across platforms and toolchains. Because svctm definitions and units vary, comparisons between systems should be done cautiously and in conjunction with vendor-specific documentation and other metrics such as await, util, and queue depth.